Skip to main content

For Dataset Publishers

To see a dataset in the Visual Dataset Browser, create a plugin instance of pl-visual-dataset. The parameters for pl-visual-dataset are also used to assign tags to files and associate metadata with tag sets such as author, academic references, and default Niivue options.

tip

This documentation is for pl-visual-dataset version 0.3.0, which currently supports .nii.gz, .nii, and .mgz volumes.

Well-Known Modes

Currently, there are three modes:

  • FreeSurfer: --mode=freesurfer-7.3.3
  • MALP-EM: --mode=malpem-1.3
  • Everything mode (no arguments)

When you use a value for --mode, pl-visual-dataset will select known output files from the specified program and associate them with tags. For example, --mode=freesurfer-7.3.3 will cause the file mri/aparc.DKTatlas+aseg.mgz to be recognized as a volume containing labels needing the FreeSurferColorLUT.txt color map.

If pl-visual-dataset is run without arguments, all supported files (NIFTI, MGZ) will be matched, but but no tags will be associated.

Advanced: Matchers and Options

To run pl-visual-dataset, provide a value for --matchers. Typically, this is done by uploading a JSON file to ChRIS then running pl-tsdircopy followed by a pl-topologicalcopy in the same feed.

Screenshot showing pl-tsdircopy

The JSON should be a list of objects which include a key-value pair along with a regular expression matching files which should be tagged with the key-value pair.

[
{
"key": "author",
"value": "Kiho Im",
"regex": "/kiho\\.nii\\.gz$"
},
{
"key": "institution",
"value": "Boston Children's Hospital",
"regex": "/(kiho|ali).*\\.nii\\.gz$"
}
]

In this example, files called kiho.nii.gz will be tagged with author=Kiho Im and files called kiho.nii.gz or ali.nii.gz will be tagged with institution=Boston Children's Hospital.

Python (and LSP/IDE auto-complate + validation) can be used to generate these JSON documents. See here for an example.

Options

Additional metadata, such as the dataset's website, academic publications, or a volume's preferred colormap can be specified for tag sets. For example, we want all files tagged with author=Kiho Im to be associated with the website https://research.childrenshospital.org/neuroim/, and all files tagged by both contrast=T2 and type=MRI to have a gray colormap.

[
{
"match": { "author": "Kiho Im" },
"options": {
"website": "https://research.childrenshospital.org/neuroim/"
}
},
{
"match": { "contrast": "T2", "type": "MRI" },
"options": {
"niivue_defaults": {
"colormap": "gray"
}
}
}
]